The Effects of After-Hours Information on Stock Prices and Trading Volume ()
1. Introduction
As information appears, such as that on increasing tariff barriers, it may cause fluctuations in the stock prices of related industries. A very active research field is attempting to discover the relationship between information and prices and trading volume, especially during the recent U.S.-China trade dispute.
This study seeks to understand the relationship between information and market performance, and contribute to the prediction of prices by information. First, this study examines the question “how does information volume affect stock prices?” Second, trading volume also needs to be investigated. To do this systemically, the impact of the quantity of information and the level of information demand on prices and trading volume are discussed, respectively. However, the strength of the information supply can predict prices, but unlike some other studies, it is irrelevant between information search volume and prices. Retail investors in the Taiwan market search external information that cannot bring about the expected price changes. Investors’ search keeps the price unchanged but brings trading volume, and so-called noise investors may exist in the market. Instead, when information appears, especially an urgent and large amount, the nature of such information will affect subsequent stock prices.
A renewed exploration is provided in two areas of information and price research. First, a control-time-lag method was proposed to construct the data. Da et al. (2011) were the first to consider Google’s search volume index as an indicator of investor attention and used it to observe stock prices. Many subsequent studies suggest that firm-specific and market-level information are correlated with prices and trade volume (Vlastakis & Markellos, 2012; Bijl et al., 2016; Li et al., 2017; Moussa et al., 2017a, 2017b). However, the related studies mostly utilize non-daily data, especially for relatively long-term monthly data (e.g., Li et al. (2017)) or weekly data (e.g., Da et al. (2011); Vlastakis & Markellos (2012); Bijl et al. (2016); Moussa et al. (2017a, 2017b)), the time period of data between the independent (information-related variables) and dependent variables (price and trading volume) overlap, and the nature of causality and simultaneity may reduce the authenticity of causal effects. To avoid time overlap, daily data should be used. Information-related variables were collected after the closing of the market on Friday, and how information affects stock prices on the next Monday was investigated. In other words, this study focuses on how the information released overnight on the weekend during the pre-opening period in the Taiwan Stock Exchange (TWSE) affects stock prices.
Second, this research expands the event study method to further examine the volume of information about the event. For example, Wang et al. (2005) analyzed a tax reduction policy event in Taiwan from August 2001 to early 2002. They defined ten points in time where relevant information appeared as event days, and observed differences in market performance before and after the event; however, the authors did not consider the effects of the difference in information volume at each point. Fortunately, this study focused on the early period of the U.S.-China trade dispute as the sample period, constructing an indicator to capture the imposing tariff information volume. During the period, only steel and aluminum goods were subject to tariffs, which triggered the investigation of related industries.
This study uses market information to predict stock prices and trading volumes. First, different from other studies, the current research finds that it is invalid to predict stock prices based on search volume. The public information search of retail investors brings trading volume, but cannot effectively determine prices. This implies that retail investors in the Taiwan stock market may be noise traders. Their search for market information or even the exchange of information with each other is invalid for stock price changes. Second, to predict stock prices, it is better to pay attention to the public information supply volume that has not yet been reflected in the market, such as after-market news and newspaper reports. The stronger the information supply, the greater the impact on stock prices. Moreover, some phenomena, such as the effects of investors’ inconsistent investment behaviors (Beaver, 1968; Kim & Verrecchia, 1991) and their attention (Moussa et al., 2017a, 2017b) on market prices or volume have also been confirmed in the Taiwan market.
The remainder of this paper is organized as follows. Section 2 presents the links between information and stock prices and trading volume. Section 3 describes the data sources, data collection methods, and empirical models. Section 4 presents the main empirical findings, and Section 5 provides some brief empirical results and implications.
2. Information, Investors, and the Market
From the emergence of market-level information to the change in stock market equilibrium, two link relations have been actively investigated. It includes 1) investors’ search activities (Da et al., 2011; Vlastakis & Markellos, 2012; Bijl et al., 2016; Li et al., 2017; Moussa et al., 2017a, 2017b) and 2) investors’ judgment and trading behaviors (Beaver, 1968; Kim & Verrecchia, 1991; Moussa et al., 2017a, 2017b), eventually causing changes in market equilibrium.
The first link between information and market changes lies in investors’ search for information. Da et al. (2011) were the first to consider Google’s search volume index as an indicator of investor attention, and confirmed that when the search attention is higher, the short-term stock price is upward. Joseph et al. (2011) also demonstrated that investment targets with higher search volumes have higher abnormal returns. However, Bijl et al. (2016) used Google’s index to predict stock returns, but the empirical results were quite different. According to Bijl et al. (2016), the sample period includes the financial crisis (2008) and the later years, which is later than the sample period in Da et al. (2011) and Joseph et al. (2011), and furthermore, the closer to the present, the more quickly information spread and integrated into the market, so the weekly data they used does not reflect the immediate effect of the information on the market, and only the subsequent downward performance can be observed.
In addition, weekly or longer data frequency may also cause time overlap between variables, causing bias due to causality or simultaneity. Since Da et al. (2011) used the Google search volume index to observe stock prices, many studies have used the index to predict the market (Vlastakis & Markellos, 2012; Bijl et al., 2016; Li et al., 2017; Moussa et al., 2017a, 2017b). Most of these studies took the index as an information demand volume indicator, and took a certain length of time as a unit to observe how it affects the market. This unit of time is usually a month (e.g., Li et al. (2017)) or week (e.g., Da et al. (2011); Vlastakis & Markellos (2012); Bijl et al. (2016); Moussa et al. (2017a, 2017b)). However, because the period of monthly or weekly data is relatively long, there may be interactions between dependent (market indicators) and independent variables (information-related variables), and causality or simultaneity problems exist between variables. Hence, this study utilized daily control-time-lag data.
The second link is investors’ trading judgment behaviors. Beaver (1968) suggested that when information appears, investors’ perceptions of whether the information is good or bad for the market lead to changes in market prices, but if investors’ opinions are divergent, trading volume will increase. On the other words, if investment judgments are divergent, the current prices may no fluctuate, but the trading volume will be pushed up. Kim and Verrecchia (1991) used a rational expectations model to theoretically identify Beaver’s intuition. The emergence of information makes investors make trading judgments and are naturally prone to occur on information-related stocks, which is also the attention effect of the information. Moussa et al. (2017a, 2017b) examined the impact of information on the market and distinguished information-related variables as information demand (information searching volume) and supply (the amount of related information news). For stock prices, they concluded the direction of the price change depends on whether the news contains positive or negative information. For stock trading volume, because the information causes investors to pay more attention to the investment target, the tendency to be traded will be greater than those that have not been noticed. Hence, they found both information supply and demand volume are influential, and the information demand factor is more important.
To understand the relationship between information and the market, and to adjust the causality and simultaneity problems, this study uses control-time-lag daily data to estimate the impact of the after-hours information over the weekend that are not reflected in the market on the Monday opening market. In general terms, a straightforward representation of the relationship between information and the market is as follows:
(1)
where
is the stock price and trading volume change,
is information supply volume,
is information demand volume,
is the international stock market information,
is a vector of other determinants of market change.
3. Data and Empirical Models
This study focuses on the Taiwan stock market, and chooses the early period of the U.S.-China trade dispute as the sample period (24/2/2018-26/3/2018). During the dispute, there was information on imposing tariffs on many commodities, but in the early stage, only steel and aluminum goods were the subjects, which allowed this study to focus on the impact of the specific industry’s information on the stock prices and trading volume of companies in related industries1. The Taiwan market is a small open market with a high degree of connection with international trade, making it suitable for observing the correlation of disputes among major international markets.
A noteworthy phenomenon was clearly observed during this period. After the Taiwan stock market closed on March 23, 2018 (Friday), the U.S. stock market, which is in another time zone opened, and the Dow Jones Industrial Average (DJIA) Index fell about 400 points on that day2. On the weekend (24-25/3, the 4th weekend in March), 28 news reports mentioned the issue of steel and aluminum tariffs between U.S. and China, and 15 news items were reported on the previous weekend3; the intensity of the information supply volume on the 4th weekend was stronger than that on the previous weekend. Comparing the performance of Taiwan stocks, the TWSE Index at the opening of the next Monday (March 26) was 40 points lower than the last Friday, and the daily trading volume was about NT $121 billion, a reduction of nearly 30 billion from the last Friday (see Figure 1). The negative after-market information about the decline in the international market and the trade dispute seemed to have affected stocks in the Taiwan market and caused a downward trend. Moreover, during the U.S.-China trade dispute period, when the negative imposing tariffs event information volume increased, but the information on international market performance was positive, both positive and negative signals spread across the market. Beaver (1968) and Kim and Verrecchia (1991) find the inconsistent investment judgments of investors may increase the stock trading volume. On the 4th weekend of March, the negative information on the trade dispute increased, and the performance of the
![]()
Figure 1. Stock markets and information supply in the 4th week of March 2018. Note: 1. Except for the index of the Taiwan stock market on March 26, which represents the index at the opening time, the rest are at the closing time. 2. The trading volume is the total daily volume of the market. Source: the amount of information is calculated by counting the number of related news reported by udndata.com; the DJIA index and the TWSE index and trading volume are from available historical data; the chart is made by the author.
international market was also negative. The consistent negative information seemed to lead investors’ judgments to be more consistent and decreased trading volume.
The empirical data include 109 companies listed on the TWSE. Each company is denoted in the sequence of numbers i (i = 1, …, 109). It covers the steel industry (30 companies) as well as other related industries, including the motor vehicle manufacturing industry (8 companies), the construction industry (49 companies), and the food products industry (22 companies).
The data were collected mainly from four sources. First, the source of information supply for steel and aluminum tariffs comes from udndata.com4. Second, the source of information demand comes from Google Trends. Third, the stock prices, trading volume, firm size, and price-to-book (P/B) ratio variables are from the Taiwan Economic Journal. Finally, the information for the U.S. stock price index comes from available past public historical data.
To understand the impact of information on prices and trading volume, the key independent variables are information-related variables, including the variables of information supply volume, information demand volume, and international stock market information. To control the time lag between independent and dependent variables, information-related data were collected for two days on the weekend (Saturday and Sunday), and the impact on the stock market at the Monday opening is observed.
Information supply volume (
). All the quantities of information reported by udndata.com in the five weekends (the news reported on Saturday and Sunday, t = 1, …, 5) was collected. First, traditional Chinese character keywords “China,” “U.S.,” “trade,” and “steel” were used to search for relevant news reported in the five weekends. Second, an analysis of articles, calculating the number of reports that specifically mentioned steel and aluminum tariffs on each weekend, and the indicator
was set as the information supply volume in the tth weekend.
Information demand volume (
). The search volume data provided by Google Trends were used to measure the information demand volume. First, traditional Chinese characters “trade” and “war” were utilized as keywords, and the sample period as the period range of searching, obtaining the search volume index for each day5. Second, the average of the index for two days of Saturday and Sunday was calculated for every week, and generating a variable,
.
International stock market information (
). Since this study analyzes U.S.-China trade dispute information, further consideration was given to the stock market performance information of these two major international markets. However, collinearity challenges must be addressed. Since the sample covers a five-week period, if the empirical model has too many time-specific variables, there may be an issue of collinearity. Because the entire period during opening to closing of the DJIA on Friday will occur after the closing time of the TWSE on Friday, it captures the complete reflection of the last Friday’s U.S. stock performance on the opening of Monday’s TWSE stock market, not just a portion of it, and it is also appropriate to the scope of the after-hours information herein. Hence, the DJIA Index performance is retained in the empirical models as an indicator of the performance of the international stock market, named
.
To observe the effect of the information on stock price and trading volume, it is necessary to set up
and
indicators. The former calculates the price change rate of each stock on Monday’s opening relative to closing on the last Friday. However, only daily data of stock trading volume can be obtained, and the ratio of the trading volume change (
) is calculated on Monday relative to last Friday.
Some control variables are also set in the models, where
is an indicator of the local overall market performance,
represents the firm size,
is the P/B ratio, and
,
, and
are industry dummies representing motor vehicle manufacturing, construction, and food product industries, respectively. All variables are presented in Table 1. OLS regression models are used for empirical research, and the complete model is presented below.
(2)
4. Empirical Results
The descriptive statistics of the data are presented in Table 2. Due to this article aims to study the relationship between information and stock prices and trading volume, first regarding the two independent variables, the average stock price change rate is positive (0.30%), with a maximum value of 8.47% and a minimum value of −3.03%. The average stock trading volume change rate is 0.42%, while the maximum and minimum values are 22.54% and −0.95%. On the other hand,
*: A. from the Taiwan Economic Journal; B. from udndata.com; C. from Google Trends; D. from available historical data.
Note: Observations = 545.
the information supply volume (the number of related news) ranges from 3 to 45, and the demand volume (ranking from 0 - 100) ranges from 1 to 77.5.
The estimated results of the OLS models are reported in Table 3. First, information search volume corresponds to stock trading volume, but is not effective for prices. Column 3 displays the price change variable (Group (a)) on two information volume indicators. As expected, because the imposing tariff information is negative for the market, more negative news volume and more negative information searching cause market prices to be more downward. In Column 5, the INMKT indicator is added. However, in Column 5 and the complete model (Column 6), among the three information-related indicators, only the information supply volume significantly affects stock prices. In contrast to the literature (e.g., Da et al. (2011); Joseph et al. (2011); Bijl et al. (2016)), the INFOD indicator has no significant impact on prices, implying that retail investors search for existing market information, which cannot be significantly reflected in stock prices in the Taiwan market. Professional investors generally have more resources and internal self-research than retail investors, and retail investors normally obtain information through the internet. This is why the literature regards internet search as a proxy for retail investors’ attention (e.g., Da et al. (2011); Dimpfl & Jank (2016)). However, in Group (b), Column 6 shows that retail investors’ attention (INFOD) significantly increases trade volume. In other words,
(a) ![]()
(b) ![]()
Table 3. Estimated results of the OLS models. (a) Stock prices change; (b) Stock trading volume change.
Note: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively. Standard errors are in parentheses.
the noise traders exist, who search for external noise information and trade, but cannot effectively affect stock prices exist in the Taiwan market.
Second, predicting the market price depends on the nature of the information that appears in the market. In Column 6, INFOS (coefficient = −0.0141) is the only information-related variable that significantly affects stock price change. As previously mentioned, the emergence of negative news has a negative impact on stock prices. The prediction of stock prices change depends on the nature of market information, which echoes the argument of Wang et al. (2005) and is also consistent with Moussa et al. (2017a, 2017b). Furthermore, in the Taiwan market, this effect is more important than that of investors’ attention represented by information searching.
Third, the influence of information-related variables on the Taiwan stock market seems to be more important than the fundamental variables of listed companies. This is also different from some studies that found company fundamentals, such as firm size, make the information differentially affect stock prices (e.g., Marshall & Walker (2002)). From the complete model (Column 6), in Group (b), in terms of trading volume change, INFOS, INFOD, and INMKT are all positive and significant, indicating that a large number of reports on market-related news, investors’ high-level search for information, and news of changes in international markets will all increase market trading activities. However, in Group (a), in addition to the differences in industry characteristics, the supply volume of news is the most important factor affecting prices. Investors assess the expected price of the stock based on market information, and the price spread with the current price makes investors willing to trade, increasing trading volume, and adjusting prices.
5. Conclusion
Although many studies have examined the impact of information search and news volume on the market, some do not consider the causality or simultaneity problem caused by time overlap between variables. The current study used the control-time-lag method to collect pre-opening information that has not yet been reflected in the Taiwan stock market during the weekend and how it affects the market. First, retail investors search for public information, and the trading volume increases but the prices cannot be changed, indicating that there may be noise traders. Second, because noise traders may search market noise, observing investors’ attention to predict stock prices is invalid. To predict the price, it should return to the essence of the information, which is good or bad for the market. Third, information-related factors are more important than companies’ fundamentals in affecting the Taiwan stock market.
In terms of follow-up research, based on this study, the following are suggested as avenues for future research: 1) some literature mentioned stock information content (e.g., Cao (2017)) or further distinguished the information into good and bad information for discussion (e.g., Depken (2001)), which is a topic for further study. 2) The dynamic process should be investigated to further distinguish how information between different industries affects each other’s industrial markets.
NOTES
1In the process of the U.S.-China trade dispute, wide ranges of goods were involved in the issue of imposing tariffs (Kim & Margalit, 2021; Chen et al., 2022). In order to avoid the interference of information noise from different goods’ tariffs, this study focuses on the early stage of the dispute, in which only steel and aluminum goods were the targets.
2Source: from available historical data.
3Source: the number of news comes from the amount of related news items in the news database udndata.com.
4udndata.com included news sources such as the United Daily News, Economic Daily News, United Evening News, and Upaper in Taiwan, collectively covering a wide range of information.
5The keywords are more specific from INFOSt than INFODt because when information demanders need relevant information, they will search the internet with intuitive keywords, which are “trade” and “war”.